We tested 6 different stable diffusion AI models on the same prompt to see how they compare. These animations were generated in Automatic111 using Deforum.
Deforum is a tool that allows users to generate AI animations, AI images and AI art with different Stable Diffusion AI models checkpoints in powered by the AUTOMATIC1111 extension.
6 Stable Diffusion AI Models Experiment 1: Whimsical Donuts
Prompt: “A whimsical scene of a dreamy donut wonderland”
- Animation Mode: 3D
- Border Mode: Wrap
- Default Motion settings
- Sampler: DPM++2M SDE Karras
- Run Seed: -1
- Max Frames: 160
- Prompts positive: high-resolution, 4K UHD, intricate, cinematic
- Prompts negative: nsfw, nude
Abyss Orange Mix
This stable diffusion model is a combination of two other models, Corneo’s 7th Heaven Mix and Abyss Orange Mix 3. It’s a good choice for those who enjoy the anime style but also want to incorporate the advanced lighting and explicit content of AOM3 without sacrificing the sharper look.
Analog Diffusion is a dreambooth model trained on a widely varied collection of analog photographs.
Arcane Diffusion is a Stable Diffusion model trained on images from the Television series “Arcane”.
The Stable Diffusion Inpainting Model is a checkpoint that uses heat diffusion properties to fill in missing or damaged parts of an image.
Anything V3 is a third-party model that features a fixed VAE model and CLIP position ID key, taken from Stable Diffusion V1.5. The VAE was updated using Kohya’s merge-vae script, and the CLIP was fixed with Arena’s stable-diffusion-model-toolkit web UI extensions.
A-Zovya Photoreal v1
aZoyva RPG Artist Tools v3
A-Zovya RPG Artist Tools v3 is a highly specialized image generation AI model designed for RPG (role-playing game) industry artists. The A-Zovya RPG stable diffusion model is made for creating high-quality landscapes, characters, and photorealistic images.
Break Domain M2150
Cetus Mix is a stable diffusion model that generates images in a flat anime style.
Counterfeit v25 25